A new Code of Practice on Transparency of AI-Generated Content has come into force under the provisions of Article 50 of the European Union’s AI Act.
The Code of Practice was developed by an expert group with a good amount of public consultation. The Code of Practice is voluntary, but compliance under the act is not, so those generating AI content or using generative AI services as part of their product have from now until December 2026 to understand their obligations and take the necessary steps to ensure compliance.
With the Artificial Intelligence (AI) banner being a very broad church, the act and code have had to be very specific about what the aims of this new work relate to. In essence, it’s about promoting transparency and building public awareness around some specific where misleading, deep-fake content is being generated.
Organisations will have a responsibility for labelling their content correctly. This will apply to how the consumer facing content appears, but also in how it is tagged and described for the purposes of searching and indexing.
At Dudley Editions, we use voice synthesis AI technology, working with our digital partners at ElevenLabs, to create personalised audiobooks based on voice samples that you choose to provide. We recently published some content that explains the process in a little more detail, based on some user feedback.
The works we use to create the audiobooks come from a mixture of different sources, but can mainly be summarised as falling into three camps:
Once a written work exceeds its period of copyright protection – typically 70 years after the author’s death – the rights to use the work become public. Several online repositories have grown up over the years providing eBook access to these types of work, such as Project Gutenberg;
Books that we have created ourselves. The founders of Dudley Editions are both writers to trade, so we’ve written some original works ourselves. In some cases, we’ve used AI prompts for chapter headings, but in those cases we’ve been clear about that; and
Work solicited directly from the public via our regular competitions on open calls for submission. Dudley Editions is a publishing house and so we publish the works of established and emerging writers.
One of the key exemptions outlined in the Code of Practice relates to the following:
Human review or editorial control for text
The disclosure obligation does not apply where the AI-generated text has undergone a process of human review or editorial control and where a natural or legal person holds editorial responsibility for the publication of the content.
The “disclosure obligation” means the need to label material that has been AI-generated, such as one of our personalised audiobooks.
As such, we don’t feel that we need to adopt the approaches, including inclusion of logos highlighting that the work has been AI-generated, because all of the work we create has gone through an editorial process which is managed and run by the founders.
While the creation of each audiobook is an automated, AI-driven event, there is still a large amount of direct human intervention with each text, book, poem or story which is made available through the Dudley Editions app and service.
In general, we’re big fans of transparency, so while we don’t think the new Code of Practice demands direct action for our products, we were keen to explore things and use the opportunity that its commencement brings to set out in more detail how we’ve thought about things and responded.
The world of generative AI is complicated and fast-moving. Codes of practice that build public understanding are a very good thing in all fields, particularly when there is the degree of complexity at play, as there is in this situation. We welcomed its arrival.
This is new legislative guidance and we’ll need to keep an eye on how the implementation goes. Perhaps our interpretation is too liberal and we’ll need to take more direct action to label that each personalised audiobook is an AI-generated product as the situation develops.
Dudley Editions has always been open about our approach and keen to develop our technologies in ways that build public trust and can realise benefits to society. Transparency is absolutely key to this digital ethics-driven approach. The new code is a timely reminder and gives us a good fresh opportunity to consider how best we describe our service and keep our audiences best informed.